Causal Inference and the Construction of Predictive Network Models in Biology - Systems Biology Concepts and Insights - page 475

Chapter 26

Causal Inference and the Construction
of Predictive Network Models in Biology

Eric E. Schadt
Department of Genetics and Genomic Sciences, Mount Sinai School of Medicine, New York City, NY 10029, USA

Chapter Outline
Introduction

499 Step 2: Reconstructing Networks using only

Expression and Metabolite Traits (Excluding DNA
Variation Data)

A Movie Analogy for Modeling Biological Systems 502

Causality as A Statistical Inference 503 509

From Assessing Causal Relationships Among Trait Pairs
to Predictive Gene Networks

Step 3: Constructing Priors Using eQTL Data 510

505 Step 4: Constructing Priors using KEGG Data 510

Building from the Bottom Up or Top Down? 505 Step 5: Constructing Networks using Expression Data,

Metabolite Data, and the Genetic and Canonical
Pathway Priors defined in Steps 3 and 4

An Integrative Genomics Approach to Constructive
Predictive Network Models 506 510

Integrating Genetic Data as a Structure prior to
Enhancing Causal Inference in the Bayesian Network
Reconstruction Process

Step 6: Comparing the Networks Constructed
in Steps 2 and 5 510

507 Networks Constructed from Human and Animal Data

Elucidate the Complexity of Disease Incorporating other 'Omics' Data as Network Priors

in the Bayesian Network Reconstruction Process

510

507 Conclusion and Future Directions 511

Illustrating the Construction of Predictive Bayesian
Networks with an Example

Unifying Bottom-Up and Top-Down Modeling
Approaches 508 512

Step 1: Identification of the URA3-Centered de novo
Biosynthesis of Pyrimidine Ribonucleotides
Subnetwork

A Need to Review How We Work Together 512

References 513

509

INTRODUCTION
The big data revolution is all around us, permeating nearly
every aspect of our lives. Electronic devices that consume
much of our attention on a daily basis enable rapid trans-
actions among individuals on an unprecedented scale,
where all of the information involved in these daily trans-
actions can be seamlessly stored in digital form, whether
the transactions involve cell phone calls, text messages,
credit card purchases, emails, or visits to the doctor's office,
in which all tests carried out are digitized and entered into
your electronic medical record ( Figure 26.1 ).
In 1984 IBM introduced the PC Junior as among the
first attempts to bring computer technology into the home.
Basic models of this first PC came with 64 kilobytes (kb) of
random access memory and a 5.25 inch floppy drive
capable of storing 360 kb of data, all at a cost of roughly
$1500 (at 2010 prices). Today, for roughly the same price,

one can purchase a laptop computer with more than a ter-
abyte of disk storage (a seven orders of magnitude increase
over the PC Junior) and a couple of gigabytes (Gb) of
random access memory (not to mention high-end graphical
processing units and high-resolution displays). The digital
universe more generally now far exceeds 1 zetabyte (that is,
21 zeros or one billion terabytes). Thus, our ability to store
and access unimaginable amounts of data has been revo-
lutionized by technological innovations that are often
observed to operate at super-Moore's Law rates.
The life and biomedical sciences have not stood on the
sidelines of this revolution. There has been an incredible
wave of new technologies in genomics such as next-

e

generation sequencing technologies [1] , sophisticated

imaging systems, and mass spectrometry-based flow

cytometry [2] enabling data to be generated at very large

scales. As a result we can monitor the expression of tens of

e